Method and apparatus for template recommendation, device, and storage medium

Abstract

The present disclosure provides a template recommendation method and apparatus, a device, and a storage medium. The method comprises: first, obtaining a feature of a multimedia material to be processed; then, determining a similarity between the feature corresponding to said multimedia material and a feature corresponding to a candidate editing template; and in response to determining that the similarity satisfies a preset matching condition, determining the candidate editing template as a target editing template corresponding to said multimedia material.

We claim: 
     
         1 . A method for template recommendation, comprising:
 obtaining a feature of a multimedia material to be processed, wherein the feature comprises a content feature of the multimedia material to be processed and a user attribute corresponding to the multimedia material to be processed, and the user attribute is obtained by analyzing a user who selects the multimedia material to be processed;   determining a similarity between the feature corresponding to the multimedia material to be processed and a feature corresponding to a candidate editing template, wherein the feature corresponding to the candidate editing template is determined based on a tag of historical multimedia works, the historical multimedia works are multimedia works edited in an editing mode applied by the candidate editing template, and the tag of the historical multimedia works comprises content features of the historical multimedia works and user attributes corresponding to the historical multimedia works; and   in response to determining that the similarity satisfies a preset matching condition, determining the candidate editing template as a target editing template corresponding to the multimedia material to be processed, wherein the target editing template is applied to edit the multimedia material to be processed into a target multimedia work in the editing mode;   wherein before determining the similarity between the feature corresponding to the multimedia material to be processed and the feature corresponding to the candidate editing template, the method further comprises:   filtering templates in a preset template recommendation pool based on attribute information of the multimedia material to be processed, to obtain a candidate editing template, wherein the attribute information comprises at least one of: material type, material amount, material duration, or material aspect ratio, and the preset template recommendation pool comprises templates determined based on at least one of: a number of historical multimedia works, a like count of historical multimedia works, or a playback count of historical multimedia works corresponding to the templates.   
     
     
         2 . The method according to  claim 1 , wherein before determining the similarity between the feature corresponding to the multimedia material to be processed and the feature corresponding to a candidate editing template, the method further comprises:
 determining the candidate editing template, and obtaining a multimedia work collection corresponding to the candidate editing template, wherein the multimedia work collection comprises more historical multimedia works than a preset number; and   determining the feature corresponding to the candidate editing template based on at least one of: content features of historical multimedia works in the multimedia work collection, or user attributes corresponding to the historical multimedia works.   
     
     
         3 . The method according to  claim 2 , wherein a template feature vector of the candidate editing template comprises a plurality of dimensions, and determining the feature corresponding to the candidate editing template based on at least one of: content features of the historical multimedia works in the multimedia work collection, or user attributes corresponding to the historical multimedia works comprises:
 determining a tag corresponding to an individual dimension of the plurality of dimensions of the template feature vector of the candidate editing template as a target tag, wherein the target tag characterizes at least one of: the content features of the historical multimedia works in the multimedia work collection, or the user attributes corresponding to the multimedia works;   determining a proportion of historical multimedia works having the target tag in the multimedia work collection corresponding to the candidate editing template;   determining a ratio of the proportion to an average proportion corresponding to the target tag as a feature value of the candidate editing template in the individual dimension corresponding to the target tag, wherein the feature value characterizes an applicability degree of the candidate editing template in terms of the individual dimension corresponding to the target tag, and the average proportion represents an average value of proportions of historical multimedia works having the target tag in multimedia work collections corresponding to respective templates in the preset template recommendation pool; and   building a template feature vector corresponding to the candidate editing template based on the feature value of the candidate editing template in the individual dimension corresponding to the target tag.   
     
     
         4 . The method according to  claim 3 , wherein before building the template feature vector corresponding to the candidate editing template based on the feature value of the candidate editing template in the individual dimension corresponding to the target tag, the method further comprises:
 determining a distribution smoothness degree of feature values of respective templates in the preset template recommendation pool in the individual dimension corresponding to the target tag, wherein the distribution smoothness degree characterizes a difference degree among the respective templates in the preset template recommendation pool in the individual dimension corresponding to the target tag; and   determining top K target tags with lower distribution smoothness degrees as recommended standard tags, wherein K is an integer greater than 0; and   wherein building a template feature vector corresponding to the candidate editing template based on the feature value of the candidate editing template in the individual dimension corresponding to the target tag comprises:   building the template feature vector corresponding to the candidate editing template based on a feature value of the candidate editing template corresponding to the recommended standard tags.   
     
     
         5 . The method according to  claim 4 , wherein obtaining a feature of the multimedia material to be processed comprises:
 determining a feature value of the multimedia material to be processed in a dimension of the recommended standard tags based on a tag of the multimedia material to be processed; and   determining a feature vector corresponding to the multimedia material to be processed based on the feature value of the multimedia material to be processed in the dimension of the recommended standard tags; and   wherein determining the similarity between the feature corresponding to the multimedia material to be processed and the feature corresponding to the candidate editing template comprises:   calculating a similarity between the feature vector corresponding to the multimedia material to be processed and the template feature vector corresponding to the candidate editing template.   
     
     
         6 . The method according to  claim 1 , wherein the method is applied on a client, and after in response to determining that the similarity satisfies the preset matching condition, determining the candidate editing template as the target editing template corresponding to the multimedia material to be processed, the method further comprises:
 editing the multimedia material to be processed in an editing mode applied by the target editing template to obtain the target multimedia work.   
     
     
         7 . The method according to  claim 6 , wherein the target editing template comprises a plurality of candidate editing templates, and before editing the multimedia material to be processed in the editing mode applied by the target editing template to obtain the target multimedia work, the method further comprises:
 presenting the plurality of candidate editing templates; and   wherein editing the multimedia material to be processed in the editing mode applied by the target editing template to obtain the target multimedia work comprises:   in response to performing selection operation on any one of the plurality of candidate editing templates, editing the multimedia material to be processed in an editing mode applied by a template corresponding to the selection operation to obtain the target multimedia work.   
     
     
         8 . An electronic device, comprising:
 a processor; and   a memory, configured to store instructions executable by the processor;   the processor configured to read the executable instructions from the memory and execute the instructions and configured to:   obtain a feature of a multimedia material to be processed, wherein the feature comprises a content feature of the multimedia material to be processed and a user attribute corresponding to the multimedia material to be processed, and the user attribute is obtained by analyzing a user who selects the multimedia material to be processed;   determine a similarity between the feature corresponding to the multimedia material to be processed and a feature corresponding to a candidate editing template, wherein the feature corresponding to the candidate editing template is determined based on a tag of historical multimedia works, the historical multimedia works are multimedia works edited in an editing mode applied by the candidate editing template, and the tag of the historical multimedia works comprises content features of the historical multimedia works and user attributes corresponding to the historical multimedia works; and   in response to determining that the similarity satisfies a preset matching condition, determine the candidate editing template as a target editing template corresponding to the multimedia material to be processed, wherein the target editing template is applied to edit the multimedia material to be processed into a target multimedia work in the editing mode;   wherein before determining the similarity between the feature corresponding to the multimedia material to be processed and the feature corresponding to the candidate editing template, the method further comprises:   filtering templates in a preset template recommendation pool based on attribute information of the multimedia material to be processed, to obtain a candidate editing template, wherein the attribute information comprises at least one of: material type, material amount, material duration, or material aspect ratio, and the preset template recommendation pool comprises templates determined based on at least one of: a number of historical multimedia works, a like count of historical multimedia works, or a playback count of historical multimedia works corresponding to the templates.   
     
     
         9 . The electronic device according to  claim 8 , wherein the processor is further configured to: before determining the similarity between the feature corresponding to the multimedia material to be processed and the feature corresponding to a candidate editing template,
 determine the candidate editing template, and obtain a multimedia work collection corresponding to the candidate editing template, wherein the multimedia work collection comprises more historical multimedia works than a preset number; and   determine the feature corresponding to the candidate editing template based on at least one of: content features of historical multimedia works in the multimedia work collection, or user attributes corresponding to the historical multimedia works.   
     
     
         10 . The electronic device according to  claim 9 , wherein a template feature vector of the candidate editing template comprises a plurality of dimensions, and the processor is configured to determine the feature corresponding to the candidate editing template based on at least one of: content features of the historical multimedia works in the multimedia work collection, or user attributes corresponding to the historical multimedia works in the following manner:
 determining a tag corresponding to an individual dimension of the plurality of dimensions of the template feature vector of the candidate editing template as a target tag, wherein the target tag characterizes at least one of: the content features of the historical multimedia works in the multimedia work collection, or the user attributes corresponding to the multimedia works;   determining a proportion of historical multimedia works having the target tag in the multimedia work collection corresponding to the candidate editing template;   determining a ratio of the proportion to an average proportion corresponding to the target tag as a feature value of the candidate editing template in the individual dimension corresponding to the target tag, wherein the feature value characterizes an applicability degree of the candidate editing template in terms of the individual dimension corresponding to the target tag, and the average proportion represents an average value of proportions of historical multimedia works having the target tag in multimedia work collections corresponding to respective templates in the preset template recommendation pool; and   building a template feature vector corresponding to the candidate editing template based on the feature value of the candidate editing template in the individual dimension corresponding to the target tag.   
     
     
         11 . The electronic device according to  claim 10 , wherein the processor is further configured to: before building the template feature vector corresponding to the candidate editing template based on the feature value of the candidate editing template in individual dimension corresponding to the target tag,
 determining a distribution smoothness degree of feature values of respective templates in the preset template recommendation pool in the individual dimension corresponding to the target tag, wherein the distribution smoothness degree characterizes a difference degree among the respective templates in the preset template recommendation pool in the individual dimension corresponding to the target tag; and   determining top K target tags with lower distribution smoothness degrees as recommended standard tags, wherein K is an integer greater than 0; and   wherein building a template feature vector corresponding to the candidate editing template based on the feature value of the candidate editing template in the individual dimension corresponding to the target tag comprises:   building the template feature vector corresponding to the candidate editing template based on a feature value of the candidate editing template corresponding to the recommended standard tags.   
     
     
         12 . The electronic device according to  claim 11 , wherein the processor is configured to obtain a feature of the multimedia material to be processed in the following manner:
 determining a feature value of the multimedia material to be processed in a dimension of the recommended standard tags based on a tag of the multimedia material to be processed; and   determining a feature vector corresponding to the multimedia material to be processed based on the feature value of the multimedia material to be processed in the dimension of the recommended standard tags; and   wherein determining the similarity between the feature corresponding to the multimedia material to be processed and the feature corresponding to the candidate editing template comprises:   calculating a similarity between the feature vector corresponding to the multimedia material to be processed and the template feature vector corresponding to the candidate editing template.   
     
     
         13 . The electronic device according to  claim 8 , wherein the processor is further configured to: after in response to determining that the similarity satisfies the preset matching condition, determining the candidate editing template as the target editing template corresponding to the multimedia material to be processed,
 edit the multimedia material to be processed in an editing mode applied by the target editing template to obtain the target multimedia work.   
     
     
         14 . The electronic device according to  claim 13 , wherein the target editing template comprises a plurality of candidate editing templates, and the processor is further configured to: before editing the multimedia material to be processed in the editing mode applied by the target editing template to obtain the target multimedia work,
 present the plurality of candidate editing templates; and   wherein editing the multimedia material to be processed in the editing mode applied by the target editing template to obtain the target multimedia work comprises:   in response to performing selection operation on any one of the plurality of candidate editing templates, editing the multimedia material to be processed in an editing mode applied by a template corresponding to the selection operation to obtain the target multimedia work.   
     
     
         15 . A non-transitory computer readable storage medium, wherein instructions are stored in the computer readable storage medium, the instructions, when running on a terminal device, causing the terminal device to perform:
 obtaining a feature of a multimedia material to be processed, wherein the feature comprises a content feature of the multimedia material to be processed and a user attribute corresponding to the multimedia material to be processed, and the user attribute is obtained by analyzing a user who selects the multimedia material to be processed;   determining a similarity between the feature corresponding to the multimedia material to be processed and a feature corresponding to a candidate editing template, wherein the feature corresponding to the candidate editing template is determined based on a tag of historical multimedia works, the historical multimedia works are multimedia works edited in an editing mode applied by the candidate editing template, and the tag of the historical multimedia works comprises content features of the historical multimedia works and user attributes corresponding to the historical multimedia works; and   in response to determining that the similarity satisfies a preset matching condition, determining the candidate editing template as a target editing template corresponding to the multimedia material to be processed, wherein the target editing template are applied to edit the multimedia material to be processed into a target multimedia work in the editing mode;   wherein before determining the similarity between the feature corresponding to the multimedia material to be processed and the feature corresponding to the candidate editing template, the method further comprises:   filtering templates in a preset template recommendation pool based on attribute information of the multimedia material to be processed, to obtain a candidate editing template, wherein the attribute information comprises at least one of: material type, material amount, material duration, or material aspect ratio, and the preset template recommendation pool comprises templates determined based on at least one of: a number of historical multimedia works, a like count of historical multimedia works, or a playback count of historical multimedia works corresponding to the templates.   
     
     
         16 . The non-transitory computer readable storage medium according to  claim 15 , wherein before determining the similarity between the feature corresponding to the multimedia material to be processed and the feature corresponding to a candidate editing template, the method further comprises:
 determining the candidate editing template, and obtaining a multimedia work collection corresponding to the candidate editing template, wherein the multimedia work collection comprises more historical multimedia works than a preset number, and   determining the feature corresponding to the candidate editing template based on at least one of: content features of historical multimedia works in the multimedia work collection, or user attributes corresponding to the historical multimedia works.   
     
     
         17 . The non-transitory computer readable storage medium according to  claim 16 , wherein a template feature vector of the candidate editing template comprises a plurality of dimensions, and determining the feature corresponding to the candidate editing template based on at least one of: content features of the historical multimedia works in the multimedia work collection, or user attributes corresponding to the historical multimedia works comprises:
 determining a tag corresponding to an individual dimension of the plurality of dimensions of the template feature vector of the candidate editing template as a target tag, wherein the target tag characterizes at least one of: the content features of the historical multimedia works in the multimedia work collection, or the user attributes corresponding to the multimedia works;   determining a proportion of historical multimedia works having the target tag in the multimedia work collection corresponding to the candidate editing template;   determining a ratio of the proportion to an average proportion corresponding to the target tag as a feature value of the candidate editing template in the individual dimension corresponding to the target tag, wherein the feature value characterizes an applicability degree of the candidate editing template in terms of the individual dimension corresponding to the target tag, and the average proportion represents an average value of proportions of historical multimedia works having the target tag in multimedia work collections corresponding to respective templates in the preset template recommendation pool; and   building a template feature vector corresponding to the candidate editing template based on the feature value of the candidate editing template in the individual dimension corresponding to the target tag.   
     
     
         18 . The non-transitory computer readable storage medium according to  claim 17 , wherein before building the template feature vector corresponding to the candidate editing template based on the feature value of the candidate editing template in the individual dimension corresponding to the target tag, the method further comprises:
 determining a distribution smoothness degree of feature values of respective templates in the preset template recommendation pool in the individual dimension corresponding to the target tag, wherein the distribution smoothness degree characterizes a difference degree among the respective templates in the preset template recommendation pool in the individual dimension corresponding to the target tag; and   determining top K target tags with lower distribution smoothness degrees as recommended standard tags, wherein K is an integer greater than 0; and   wherein building a template feature vector corresponding to the candidate editing template based on the feature value of the candidate editing template in the individual dimension corresponding to the target tag comprises:   building the template feature vector corresponding to the candidate editing template based on a feature value of the candidate editing template corresponding to the recommended standard tags.